Whispered speech differs from normal speech by having no periodic excitation or harmonic structure, has formant frequencies that are shifted higher than those of normal speech, has a spectral slope that is flatter than the spectral slope of normal speech, has a longer duration than normal speech and exhibits a lower energy level.
Due to these differences it is hard to detect whispered speech and usage of normal speech detection algorithms is inefficient.
There is a growing need to detect whispered speech.